MIT Media Lab professor Ramesh Raskar discusses a vision of AI's future centered on billions of personal agents that users can train and control, rather than the pursuit of . In conversation with journalist Nina Gregory, Raskar argues that small, specialized models running on phones and personal devices could handle everyday tasks more effectively than pursuing AGI.
The discussion explores edge AI, decentralized intelligence, and the infrastructure needed for what Raskar calls an " web," including identity verification, testing and governance frameworks. He also introduces Project NANDA and argues that making an AI agent safe involves more than just securing its underlying model—it requires broader system-level considerations.
This is an AI-written summary. The original reporting belongs to Imagination in Action. How we write these · Spot a mistake?